* Expand histogram memory dynamically to prevent large allocations for large tree depths (e.g. > 15)
* Remove GPU memory allocation messages. These are misleading as a large number of allocations are now dynamic.
* Fix appveyor R test
* Added finding quantiles on GPU.
- this includes datasets where weights are assigned to data rows
- as the quantiles found by the new algorithm are not the same
as those found by the old one, test thresholds in
tests/python-gpu/test_gpu_updaters.py have been adjusted.
* Adjustments and improved testing for finding quantiles on the GPU.
- added C++ tests for the DeviceSketch() function
- reduced one of the thresholds in test_gpu_updaters.py
- adjusted the cuts found by the find_cuts_k kernel
* Upgrading to NCCL2
* Part - II of NCCL2 upgradation
- Doc updates to build with nccl2
- Dockerfile.gpu update for a correct CI build with nccl2
- Updated FindNccl package to have env-var NCCL_ROOT to take precedence
* Upgrading to v9.2 for CI workflow, since it has the nccl2 binaries available
* Added NCCL2 license + copy the nccl binaries into /usr location for the FindNccl module to find
* Set LD_LIBRARY_PATH variable to pick nccl2 binary at runtime
* Need the nccl2 library download instructions inside Dockerfile.release as well
* Use NCCL2 as a static library
* Refactor to allow for custom regularisation methods
* Implement compositional SplitEvaluator framework
* Fixed segfault when no monotone_constraints are supplied.
* Change pid to parentID
* test_monotone_constraints.py now passes
* Refactor ColMaker and DistColMaker to use SplitEvaluator
* Performance optimisation when no monotone_constraints specified
* Fix linter messages
* Fix a few more linter errors
* Update the amalgamation
* Add bounds check
* Add check for leaf node
* Fix linter error in param.h
* Fix clang-tidy errors on CI
* Fix incorrect function name
* Fix clang-tidy error in updater_fast_hist.cc
* Enable SSE2 for Win32 R MinGW
Addresses https://github.com/dmlc/xgboost/pull/3335#issuecomment-400535752
* Add contributor
* Use sparse page as singular CSR matrix representation
* Simplify dmatrix methods
* Reduce statefullness of batch iterators
* BREAKING CHANGE: Remove prob_buffer_row parameter. Users are instead recommended to sample their dataset as a preprocessing step before using XGBoost.
* GPU binning and compression.
- binning and index compression are done inside the DeviceShard constructor
- in case of a DMatrix with multiple row batches, it is first converted into a single row batch
* Increase precision of bst_float values in tree dumps
* Increase precision of bst_float values in tree dumps
* Fix lint error and switch precision to right float variable
* Fix clang-tidy error
* Multi-GPU HostDeviceVector.
- HostDeviceVector instances can now span multiple devices, defined by GPUSet struct
- the interface of HostDeviceVector has been modified accordingly
- GPU objective functions are now multi-GPU
- GPU predicting from cache is now multi-GPU
- avoiding omp_set_num_threads() calls
- other minor changes
* Extended monotonic constraints support to 'hist' tree method.
* Added monotonic constraints tests.
* Fix the signature of NoConstraint::CalcSplitGain()
* Document monotonic constraint support in 'hist'
* Update signature of Update to account for latest refactor
* Replaced std::vector-based interfaces with HostDeviceVector-based interfaces.
- replacement was performed in the learner, boosters, predictors,
updaters, and objective functions
- only interfaces used in training were replaced;
interfaces like PredictInstance() still use std::vector
- refactoring necessary for replacement of interfaces was also performed,
such as using HostDeviceVector in prediction cache
* HostDeviceVector-based interfaces for custom objective function example plugin.
* Added GPU objective function and no-copy interface.
- xgboost::HostDeviceVector<T> syncs automatically between host and device
- no-copy interfaces have been added
- default implementations just sync the data to host
and call the implementations with std::vector
- GPU objective function, predictor, histogram updater process data
directly on GPU
* Fix#2905
* Fix gpu_exact test failures
* Fix bug in GPU prediction where multiple calls to batch prediction can produce incorrect results
* Fix GPU documentation formatting
- Implement colsampling, subsampling for gpu_hist_experimental
- Optimised multi-GPU implementation for gpu_hist_experimental
- Make nccl optional
- Add Volta architecture flag
- Optimise RegLossObj
- Add timing utilities for debug verbose mode
- Bump required cuda version to 8.0
* Fatal error if GPU algorithm selected without GPU support compiled
* Resolve type conversion warnings
* Fix gpu unit test failure
* Fix compressed iterator edge case
* Fix python unit test failures due to flake8 update on pip
* [R] MSVC compatibility
* [GPU] allow seed in BernoulliRng up to size_t and scale to uint32_t
* R package build with cmake and CUDA
* R package CUDA build fixes and cleanups
* always export the R package native initialization routine on windows
* update the install instructions doc
* fix lint
* use static_cast directly to set BernoulliRng seed
* [R] demo for GPU accelerated algorithm
* tidy up the R package cmake stuff
* R pack cmake: installs main dependency packages if needed
* [R] version bump in DESCRIPTION
* update NEWS
* added short missing/sparse values explanations to FAQ
* Removal of redundant code/files.
* Removal of exact namespace in GPU plugin
* Revert double precision histograms to single precision for performance on Maxwell/Kepler
It has been reported that new parallel algorithm (#2493) results in excessive
message usage (see issue #2326). Until issues are resolved, XGBoost should use
the old parallel algorithm by default. The user would have to specify
`enable_feature_grouping=1` manually to enable the new algorithm.
* Patch to improve multithreaded performance scaling
Change parallel strategy for histogram construction.
Instead of partitioning data rows among multiple threads, partition feature
columns instead. Useful heuristics for assigning partitions have been adopted
from LightGBM project.
* Add missing header to satisfy MSVC
* Restore max_bin and related parameters to TrainParam
* Fix lint error
* inline functions do not require static keyword
* Feature grouping algorithm accepting FastHistParam
Feature grouping algorithm accepts many parameters (3+), and it gets annoying to
pass them one by one. Instead, simply pass the reference to FastHistParam. The
definition of FastHistParam has been moved to a separate header file to
accomodate this change.
* Bugfix 1: Fix segfault in multithreaded ApplySplitSparseData()
When there are more threads than rows in rowset, some threads end up
with empty ranges, causing them to crash. (iend - 1 needs to be
accessible as part of algorithm)
Fix: run only those threads with nonempty ranges.
* Add regression test for Bugfix 1
* Moving python_omp_test to existing python test group
Turns out you don't need to set "OMP_NUM_THREADS" to enable
multithreading. Just add nthread parameter.
* Bugfix 2: Fix corner case of ApplySplitSparseData() for categorical feature
When split value is less than all cut points, split_cond is set
incorrectly.
Fix: set split_cond = -1 to indicate this scenario
* Bugfix 3: Initialize data layout indicator before using it
data_layout_ is accessed before being set; this variable determines
whether feature 0 is included in feat_set.
Fix: re-order code in InitData() to initialize data_layout_ first
* Adding regression test for Bugfix 2
Unfortunately, no regression test for Bugfix 3, as there is no
way to deterministically assign value to an uninitialized variable.
* Add UpdatePredictionCache() option to updaters
Some updaters (e.g. fast_hist) has enough information to quickly compute
prediction cache for the training data. Each updater may override
UpdaterPredictionCache() method to update the prediction cache. Note: this
trick does not apply to validation data.
* Respond to code review
* Disable some debug messages by default
* Document UpdatePredictionCache() interface
* Remove base_margin logic from UpdatePredictionCache() implementation
* Do not take pointer to cfg, as reference may get stale
* Improve multi-threaded performance
* Use columnwise accessor to accelerate ApplySplit() step,
with support for a compressed representation
* Parallel sort for evaluation step
* Inline BuildHist() function
* Cache gradient pairs when building histograms in BuildHist()
* Add missing #if macro
* Respond to code review
* Use wrapper to enable parallel sort on Linux
* Fix C++ compatibility issues
* MSVC doesn't support unsigned in OpenMP loops
* gcc 4.6 doesn't support using keyword
* Fix lint issues
* Respond to code review
* Fix bug in ApplySplitSparseData()
* Attempting to read beyond the end of a sparse column
* Mishandling the case where an entire range of rows have missing values
* Fix training continuation bug
Disable UpdatePredictionCache() in the first iteration. This way, we can
accomodate the scenario where we build off of an existing (nonempty) ensemble.
* Add regression test for fast_hist
* Respond to code review
* Add back old version of ApplySplitSparseData
As discussed in issue #1978, tree_method=hist ignores the parameter
param.num_roots; it simply assumes that the tree has only one root. In
particular, when InitData() method initializes row_set_collection_, it simply
assigns all rows to node 0, the value that's hard-coded.
For now, the updater will simply fail when num_roots exceeds 1. I will revise
the updater soon to support multiple roots.
